The hype around the success of mobile money in Kenya has been growing as mobile payments develop both there and worldwide. This week’s Economist cites a figure that 43% of Kenyan GDP is being channelled through M-Pesa each year, attributing the statistic to Safaricom itself. The figure has been rising from 31% last year, which was cited by both The Economist and the Financial Times. In August 2013, GSM Association released an infographic on “The Kenyan Journey to Digital Financial Inclusion”, which also used the 31% figure. The World Bank, CGAP, AFI and others have also used or cited such measures of progress in this field.
Despite the rapid rate of adoption of mobile money in Kenya, these figures raised my statistical alarm bells! 43% of GDP is an incredible figure when the maximum amount that an individual can transact in a day is Kshs 140,000 (or approximately US$1600) and the average transaction size is much smaller, actually Kshs2,634 (US$30) for Februrary 2014. Indeed, a trip to downtown Nairobi would soon seem to disabuse anyone of the idea that well over a third of transactions by number or value are being undertaken through it – cash is definitely still king.
A further problem has been the use of the number of reported connections to suggest that this is the number of people actually using the service. Let me explain both of the problems.
Problem I: Saying that 43% of GDP was transacted through mobile money is hugely misleading.
The figure is the result of dividing the volume of transactions by the figure for GDP. This is a highly misleading way of presenting this data.
A trip to the Central Bank of Kenya’s excellent data portal reveals that the value of transactions through mobile money in February 2014 was Kshs173bn. This represents the aggregated flows into and out of mobile money accounts. Just as paying for any good or service through a bank account creates both a debit to the paying party’s account and a credit to the seller’s account, this figure adds up both of these figures to give a total transactions figure of funds flowing through mobile money. That is both deposits into and withdrawals from mobile money accounts are included.
GDP on the other hand is a measure of the value of goods and services produced in an economy. It does not represent the amount of money that had to flow through the economy to pay for these goods and services, i.e. the payments made and money circulating in the economy. GDP values goods and services only once. Comparing the GDP figure with the transactions figure therefore compares figures that are not conceptually equivalent to each other.
While it would not be entirely wrong to say that “the value of transactions through mobile money is equivalent to 43% of the value of GDP” it is still very misleading because this subtle difference is not one that many people would likely understand.
To make a really robust comparison of the extent to which mobile money is making an impression on the payments landscape it is necessary to compare this volume of payments made through mobile money with a figure which represents all the payments and transfers made in the whole economy. That is of course rather tricky as such a figure needs to include the turnover of cash in a highly cash based economy like Kenya’s. And just think of how many times cash might change hands without producing any GDP. It is not simply that payments are twice the level of GDP because everything is a credit or debit for someone, but payments are also made that do not directly produce GDP, for example, any type of transfer payment such as government taxes and revenues – which in themselves are huge.
In fact, some rummaging around on the Central Bank’s website quickly produces some alternative figures which give more accurate – if rather less dramatic – comparisons:
- The figure for payments through the Kenya Electronic Payment and Settlement System (the system through which Banks settle their accounts with one another) for the same month is Kshs1,853bn. Comparing this with mobile money suggests it is facilitating transactions which are some 9% of the value facilitated by the Central Bank’s real time gross settlement system. This seems respectable given that mobile money is dealing mainly with relatively poor individuals and banks are still settling payments between businesses.
- Alternatively you could compare it to the amount being put through Kenya’s automated bank clearing which according to the website was Kshs240bn. 72% is really quite a respectable comparison for electronic payments in Kenya.
- Another approach is as follows: uniquely and due to historical issues regarding data collection, the data on Kenya’s PostBank flows of deposits and withdrawals are published in the CBK Statistical Bulletin (June 2013). These flows are not figures that commercial banks publish as they only publish stocks of deposits – not what flows across their counters. For 2013 they averaged about Kshs 8bn deposits and Kshs 8bn withdrawals per month – i.e. approx. 16bn shs of transactions in a month. Hence mobile money is currently managing payments just over 10 times that of PostBank’s transactions flow. Given PostBank is one of the smaller banks, it is therefore possible to suggest that mobile money is handling transactions roughly equivalent to the transactions one of the larger commercial banks.
It is clear then that these alternative more robust comparisons, which compare flows of transactions through mobile money with other flows of payments, put the figures in a rather more sober light while reflecting a more realistic view of the progress of mobile money payments in the payments landscape. Fortunately, having pointed this out to GSMA, they are in the process of revising their infographic and will be publishing a revised one shortly and explaining to the industry why they are no longer going to use this comparison with GDP. A letter to the The Economist last year got the response that “we don’t believe it is an apples-and-oranges comparison” and, when queried on this, that “it had not been our intention to imply that GDP was the same as the volume of payments. It’s just one way of putting M-PESA transactions in some sort of context”.
Problem II: In terms of numbers of users, there is a further mistake frequently being made – a confusion of customers with the number of unique “mobile money users”.
Kenyan mobile money is reported as having 26m “customers” (CBK website). The data collected from the providers simply reports the total number of registrations and calls them customers – it does not therefore take account of multiple registrations and hence is not a figure for unique customers. The mistake is then to divide this by the population (i.e. adult population) to suggest a proportion of the population who are “users”. This produces highly erroneous figures. Indeed with an approximate population of 43m of whom approximately half are under the age of 18, this suggests usage rates of over 100%.
The latest figure from the FinAccess 2013 nationally representative survey gives a figure of mobile money registrations at 62% of the adult population.
This is a relatively new field and clearly there is some need for data standards to be developed. Kenya’s progress is indeed impressive and Safaricom has a stake in making that apparent. Having recently reviewed Morten Jerven’s very interesting and disconcerting book on the political economy of development statistics it is less surprising to consider that there are interests at stake regarding these types of numbers and why hype might arise. He also points out that statistical offices need investment and that analysts must trace the provenance of their data in order to understand its weaknesses before presenting it – all good lessons for this case too!